virtualenv is a tool used to create an isolated workspace for a Python application. It has various advantages such as the ability to install modules locally, export a working environment, and execute a Python program in that environment.
A virtual environment is a directory into which some binaries and shell scripts are installed. The binaries include
python for executing scripts and
pip for installing other modules within the environment. There are also shell scripts (one for bash, csh, and fish) to activate the environment. Essentially, a virtual environment mimics a full system install of Python and all of the desired modules without interfering with any system on which the application might run.
Python 3.3+ comes with a tool called
pyvenv and an API called venv for extending the native implementation. For applications that require an older version of Python,
virtualenv must be used.
- Python 3.3+:
- Python 3:
- Python 2:
All three tools use a similar workflow.
virtualenv to create the virtual environment within your project directory. Be sure to exclude the venv directory from version control--a copy of
pip freeze will be enough to rebuild it.
This tool is provided by(3.3+).
$ pyvenv venv
virtualenv for Python 3, available in .
$ virtualenv venv
virtualenv2 for Python 2, available in .
$ virtualenv2 venv
Use one of the provided shell scripts to activate and deactivate the environment. This example assumes bash is used.
$ source venv/bin/activate (venv) $ which python (venv) $ deactivate # this is a shell function provided by bin/activate $ which python
Once inside the virtual environment, modules can be installed and scripts can be run as normal:
(venv) $ python -c 'import requests; print("this will not display because requests is not installed")' (venv) $ pip install requests (venv) $ python -c 'import requests; print("requests is now installed in the virtual environment")'
The binary versions depend on which virtual environment tool was used. For instance, the
python command used in the Python 2 example points to
bin/python2.7, while the one in the
pyvenv example points to
One major difference between
virtualenv is that the former uses the system's Python binary by default:
$ ls -l pyvenv/bin/python3.5 lrwxrwxrwx 1 foo foo 7 Jun 3 19:57 pyvenv/bin/python3.5 -> /usr/bin/python3
virtualenv tool uses a separate Python binary in the environment directory:
$ ls -l venv3/bin/python3.5 lrwxrwxrwx 1 foo foo 7 Jun 3 19:58 venv3/bin/python3.5 -> python3
virtualenvwrapper allows more natural command line interaction with your virtualenvs by exposing several useful commands to create, activate and remove virtualenvs. This package is a wrapper for bothand .
Now add the following lines to your
$ export WORKON_HOME=~/.virtualenvs $ source /usr/bin/virtualenvwrapper.sh
If you are not using python3 by default (check the output of
$ python --version) you also need to add the following line to your
~/.bashrc prior sourcing the
virtualenvwrapper.sh script. The current version of the
virtualenvwrapper-python package only works with python3. It can create python2 virtualenvs fine though.
Re-open your console and create the
$ mkdir $WORKON_HOME
The main information source on virtualenvwrapper usage (and extension capability) is Doug Hellmann's page.
- Create the virtualenv:
$ mkvirtualenv -p /usr/bin/python2.7 my_env
- Activate the virtualenv:
$ workon my_env
- Install some package inside the virtualenv (say, Django):
(my_env) $ pip install django
- Do your things
- Leave the virtualenv:
(my_env) $ deactivate